All measurements, although taken as accurately as possible, are subjected to uncertainty. So the analysis of errors and uncertainty is crucial in all applications since such errors need to be estimated and, when possible, reduced. In particular, when gas turbine mathematical models based on the processing of field measurements (such as the Gas Path Analysis models) are used, the evaluation of measurement reliability is a key point. In fact, it has been demonstrated that these kinds of techniques are sensitive to measurement errors: thus, tools for field data processing to evaluate the presence of the so-called outliers are advisable. In this paper, some statistical methodologies for the assessment of the reliability of the measurements taken on a gas turbine are presented. The methodologies, taken from literature and used for historical measurements, are discussed. Moreover, a new methodology, based on a modified t-Student distribution, is proposed.
On an extremal property of Doob’s class
